The broad objective of the research project is to develop improved SAR models. The general hypothesis that the authors will test is that systematic application of research criteria, procedures and methods will lead to the creation of better SAR models. In particular, the authors propose to: a) develop criteria concerning database size and features that will improve the predictive ability of the resulting SAR models, b) determine the relative effectiveness of various processes used for model building, c) research the best procedure for estimating the future prediction accuracy of an SAR model and d) characterize the power of combined SAR models or SAR model results from several model building processes. The resulting optimized SAR models with relevant descriptors of their accuracy and reliability can be used to screen untested chemicals in health risk assessment and/or to reduce or replace animal testing for certain biological endpoints. The SAR models can also help in clarifying the mechanism of the biological phenomena at the molecular level.